A compound type avalanche prevention system for a tunnel entrance in an alpine valley region

Abstract

The present application belongs to the technical field of avalanche disaster prevention, and discloses a compound avalanche prevention system for a tunnel entrance in an alpine and canyon area, which solves the technical problems of low early warning accuracy, insufficient protection capability, delayed emergency response and poor module linkage of the existing avalanche prevention system for a tunnel entrance in an alpine and canyon area. The system comprises a multi-source monitoring and early warning module, a hierarchical buffer protection module and an intelligent emergency escape module in linkage and cooperation. The multi-source monitoring and early warning module comprises a distributed sensor unit, an edge computing node and a remote monitoring center. The hierarchical buffer protection module is arranged along the avalanche movement path from top to bottom and comprises a first diversion structure, a second blocking structure and a third buffer structure. The intelligent emergency escape module comprises an escape passage body, an emergency guiding device and a life support unit. The present application is suitable for the special topographic conditions of alpine and canyon areas, can significantly improve the avalanche disaster prevention and control capability during the construction and operation periods of tunnel projects, and reduce personnel casualties and property losses.

Description

Technical Field

[0001] This invention belongs to the field of avalanche disaster prevention and control technology, and in particular relates to a composite avalanche prevention and control system for tunnel entrances in high mountain and canyon areas. Background Technology

[0002] Tunnel projects in high mountain and canyon areas are highly susceptible to avalanche hazards due to their undulating terrain and complex weather conditions. Avalanches in these areas are characterized by their large scale, suddenness, and rapid impact, which not only severely impacts construction progress but can also cause significant casualties and property damage. Existing avalanche prevention measures mainly include single monitoring equipment, fixed snow barriers, and extended tunnels, but these have several shortcomings: ① Single monitoring methods result in low early warning accuracy, making it difficult to capture the entire process of avalanche initiation and movement; ② Protective structures are mostly rigid designs with limited impact resistance and are difficult to fully cover due to site space limitations; ③ Emergency escape facilities lack intelligent guidance and protection functions, resulting in low evacuation efficiency; ④ Monitoring, protection, and emergency systems are independent and lack coordination, failing to form a closed-loop prevention and control system.

[0003] For example, a tunnel exit is located in a basin surrounded by mountains on three sides, with an altitude of 3478–4920m and a relative height difference of 1442m. Winter snowfall is heavy, wind speeds are high, and avalanches are frequent and large-scale. While existing snow barriers can block some snow, they are easily breached in the face of extreme avalanches. Although the monitoring system includes weather stations and vibration sensors, data fusion is insufficient, resulting in delayed early warning responses. The escape routes are fixed structures, lacking dynamic guidance and emergency protection, making it difficult to meet emergency evacuation needs.

[0004] Therefore, there is an urgent need to develop a composite avalanche prevention system that is adapted to the special environment of high mountains and canyons, with multi-module collaboration and a high degree of intelligence. Summary of the Invention

[0005] The purpose of this invention is to provide a composite avalanche prevention system for tunnel entrances in high mountain and canyon areas, which effectively solves the technical problems of low early warning accuracy, insufficient protection capability, delayed emergency response, and poor module linkage in existing avalanche prevention systems for tunnel entrances in high mountain and canyon areas.

[0006] To address the aforementioned technical problems, the present invention employs the following technical solution: a composite avalanche prevention system for tunnel entrances in high mountain and canyon areas, comprising a multi-source monitoring and early warning module, a tiered buffer protection module, and an intelligent emergency escape module; the multi-source monitoring and early warning module includes distributed sensor units, edge computing nodes, and a remote monitoring center; the distributed sensor units are deployed in three levels—avalanche source area, flow area, and accumulation area—around the tunnel entrance and in the construction area; the edge computing nodes have built-in AI early warning models and are deployed at the tunnel site; the tiered buffer protection module is deployed from top to bottom along the avalanche movement path, including a primary flow guiding structure, a secondary blocking structure, and a tertiary buffer structure; the primary flow guiding structure guides the avalanche to flow towards a pre-set channel, the secondary blocking structure blocks the main snow mass, and the tertiary buffer structure weakens residual impact; the intelligent emergency escape module includes an escape channel body, an emergency guidance device, and a life support unit; the escape channel body connects the construction camp and the safe area of ​​the tunnel opening, the emergency guidance device indicates the optimal escape route in real time, and the life support unit provides emergency supplies.

[0007] Furthermore, the distributed sensor unit includes: a meteorological sensor array: deployed in the avalanche source area with a spacing of ≤500m; a seismic motion sensor array: deployed in the flow area, forming a seismic network matrix with a spacing of 200-300m; a camera array: with a field of view covering the avalanche initiation area, movement path, and accumulation area of ​​the target slope; and a snow depth sensor array: deployed in key sections of the avalanche source area with a spacing of 100-150m. The key sections refer to slope sections within the avalanche source area with a slope range of 28° to 60°, possessing snow accumulation characteristics and prone to snow sliding initiation, including confluence channel sections, steep slope transition sections, ridge pass sections, and concave snow accumulation sections on the slope.

[0008] Furthermore, the multi-source monitoring and early warning module also includes a sensor adaptive calibration unit. The sensor adaptive calibration unit has a built-in environmental compensation algorithm that automatically calibrates the data accuracy of meteorological sensors, seismic sensors, and snow depth sensors based on environmental parameters at an altitude of 3478–4920 m.

[0009] Furthermore, the AI ​​early warning model, based on real-time monitoring of snow depth, wind speed, wind direction, temperature, seismic signal amplitude, snow movement image data, as well as terrain slope, altitude, and historical avalanche data, outputs three-level early warning signals and corresponding early warning response times and intelligent emergency escape module linkage control commands. The three-level early warning signals include: low-level warning: snow depth 30-50cm, wind speed 10-15m / s, seismic signal amplitude <0.1g, early warning response time ≤20s; medium-level warning: snow depth 50-60cm, wind speed 15-25m / s, seismic signal amplitude 0.1-0.3g, early warning response time ≤15s; high-level warning: snow depth >60cm, wind speed >25m / s, seismic signal amplitude >0.3g, early warning response time ≤10s.

[0010] Furthermore, the specific parameters of each structure in the graded buffer protection module are as follows: The primary diversion structure is an arc-shaped snow guide embankment, the surface of which is covered with a 100% solids content sprayed aromatic pure polyurea wear-resistant coating with a coating thickness ≥5mm. The angle between the axis of the arc-shaped snow guide embankment and the main avalanche flow direction is ≤30°, the radius of curvature is 50-80m, the height is 3-5m, and the length is 50-100m; The secondary barrier structure is a graded snow barrier dam, the total height of which is 11-15m, the dam length is 132-150m, the top width is 5-6m, and the dam slope ratio is 1:2.3-1:2.5; It is divided into a lower rigid dam body and an upper flexible protection section along the height of the dam body. The lower rigid dam body is constructed by layering and compacting tunnel spoil, and the outer side is reinforced concrete facing; the upper flexible protection section is a wire mesh protective net; The tertiary buffer structure is a foam concrete buffer layer and an energy dissipation mound array.

[0011] Furthermore, the secondary barrier structure also includes an adaptive adjustment component, which dynamically adjusts based on real-time data from the multi-source monitoring and early warning module: when the avalanche velocity is detected to be >30m / s and the snow thickness is detected to be >3m, the inclination angle of the flexible protection section is adjusted to 15° to 20° to enhance the impact buffering effect; when the wind direction shift is detected to be >10°, the arc-shaped snow guide embankment of the primary guide structure is finely adjusted in curvature by an electric push rod to maintain the guide angle ≤30°.

[0012] Furthermore, the specific parameters of the intelligent emergency escape module are as follows: the escape passage body is a steel corrugated pipe structure with an inner diameter ≥2m and a wall thickness ≥12mm. The inner wall of the passage is covered with a polyurethane insulation layer and a modified ultra-high molecular weight polyethylene impact-resistant lining. The passage length is ≤170m, and LED emergency lighting fixtures are installed along the route. The emergency guidance device includes an audible and visual alarm and a dynamic indicator. When the warning is activated, the audible and visual alarm automatically turns on, and the dynamic indicator receives signals from the multi-source monitoring and warning module to display the optimal escape route in real time. Life support unit: one set is set every 10-15m, including a portable oxygen supply device, a first aid kit, and a Beidou-3 satellite communication terminal.

[0013] Furthermore, the sensor adaptive calibration unit also includes a fault self-diagnosis function: when the sensor data deviation exceeds the calibration threshold three times consecutively, the faulty sensor is automatically marked and a maintenance prompt is sent to the remote monitoring center; the monitoring area where the faulty sensor is located automatically uses the improved ordinary Kriging interpolation algorithm to complete the data, selects effective sensors within the corresponding radius with the fault point as the center, calculates the complete value by solving the weight coefficient through spherical variogram modeling, and combines time series trend correction and multi-model redundancy mechanism to ensure that the data completion error is ≤10%.

[0014] Furthermore, the AI ​​early warning model is an avalanche risk classification and prediction model based on multi-source data fusion. It is a fusion model of LightGBM and LSTM neural networks, which includes a multi-source heterogeneous data preprocessing layer, a snow accumulation stability coefficient calculation unit, a wind speed change feature extraction unit, a vibration signal wavelet transform feature extraction unit, an environmental parameter compensation unit, a threshold screening and machine learning classification fusion judgment layer, and an early warning output layer.

[0015] Compared with the prior art, the beneficial technical effects of the present invention are: (1) The present invention integrates multi-source sensor data and introduces an AI early warning model, which effectively improves the accuracy of avalanche early warning, and the early warning response time is ≤30s, solving the problems of single monitoring methods and delayed early warning in the existing system.

[0016] (2) The present invention adopts a graded buffer protection module design to solve the site limitation problem caused by the high mountain and canyon terrain; through the synergistic effect of primary diversion, secondary blocking and tertiary buffer, the impact energy of avalanches is effectively weakened and the protection capability is improved.

[0017] (3) The intelligent emergency escape module of the present invention has dynamic guidance and life protection functions, improves evacuation efficiency, and effectively reduces the risk of casualties.

[0018] (4) This invention achieves closed-loop prevention and control throughout the entire process through multi-module linkage and adaptive regulation, adapts to different scenario requirements during the construction and operation periods, and can be widely applied to transportation engineering in high-altitude avalanche-prone areas. Attached Figure Description

[0019] Figure 1 This is a flowchart of the multi-source monitoring and early warning module of the present invention.

[0020] Figure 2 This is a graph showing the relationship between the sensor deployment spacing and the early warning response time of the present invention.

[0021] Figure 3 This is a diagram showing the adaptation relationship between the height of the snow barrier and the scale of an avalanche according to the present invention. Detailed Implementation

[0022] Example 1: This example provides a composite avalanche prevention system for tunnel entrances in high-altitude canyon areas, applied to the exit of a tunnel in XZ. The tunnel entrance is at an altitude of 3478m, with surrounding mountain slopes ranging from 28° to 60°. The peak avalanche season is from December to May of the following year. The composite avalanche prevention system includes a multi-source monitoring and early warning module, a tiered buffer protection module, and an intelligent emergency escape module, all linked and coordinated through a data transmission network. In this example, the data transmission network is a dual-backup network with both wireless and wired transmission. The wireless transmission uses 5G and BeiDou satellite dual-mode communication, while the wired transmission uses single-mode fiber optic cable. The automatic switching time between the dual-backup modes is ≤1s, achieving a closed-loop prevention and control system covering avalanche monitoring, early warning, protection, and escape.

[0023] The multi-source monitoring and early warning module includes a distributed sensor unit, an edge computing node, a remote monitoring center, and a sensor adaptive calibration unit. (1) The distributed sensor unit is deployed in three levels: avalanche source area, circulation area, and accumulation area, around the tunnel entrance and in the construction area. (2) The edge computing node is deployed near the tunnel site and has an AI early warning model built in. It processes the sensor data in real time, extracts key parameters such as snow stability characteristics and movement speed, and outputs early warning signals. (3) The remote monitoring center is set up in the project department and is linked in real time with the project department's emergency command room. It receives early warning signals and displays the avalanche movement status, and issues control commands to the graded buffer protection module and the intelligent emergency escape module. (4) The sensor adaptive calibration unit has a built-in environmental compensation algorithm and fault self-diagnosis function, supports the issuance of remote calibration commands, and leaves a full record of the calibration process. The data traceability period is ≥1 year.

[0024] like Figure 1As shown, the workflow of the multi-source monitoring and early warning module is as follows: First, multi-source data acquisition is performed to obtain various monitoring data such as meteorological data, seismic motion data, snow depth data, and video images. Then, the acquired heterogeneous multi-source data is fused to achieve spatiotemporal alignment and standardization of different types of data. After fusion, data integrity verification is performed to remove outliers and fill in missing data. After verification, risk assessment modeling is performed to extract key features such as snow stability, sudden wind speed changes, and seismic amplitude and input them into the AI ​​early warning model. The risk level is determined based on the model output. According to the determined risk level, the corresponding level of early warning information is issued, and the intelligent emergency escape module is linked in tandem. After the early warning is issued, trend monitoring and tracking are continuously performed to update the avalanche development status and risk level in real time until the avalanche risk is eliminated or the emergency response process ends.

[0025] The environmental quantities compensated by the environmental compensation algorithm in the sensor adaptive calibration unit include temperature, humidity, atmospheric pressure, high-altitude low-pressure, electromagnetic noise, light radiation, sensor temperature drift, and long-term drift. The environmental compensation algorithm adopts a combination of temperature drift lookup table method, piecewise linear calibration formula, polynomial humidity fitting, air pressure-altitude model correction, dark current correction, background subtraction, adaptive filtering, and baseline zeroing. Based on high-altitude environmental parameters (temperature -20 to 4.5℃, humidity 57% to 76%) at altitudes of 3478 to 4920 m, it automatically calibrates the sensor data accuracy every 24 hours: the meteorological sensor error after calibration is ≤ ±0.05 m / s and ±0.5℃; the snow depth sensor calibration uses laser ranging comparison method, and the error after calibration is ≤ ±1 cm; the seismic motion sensor compares with the reference vibration signal, and the sensitivity deviation after calibration is ≤ 5%.

[0026] The self-diagnosis function of the sensor adaptive calibration unit refers to the following: when the sensor data deviation exceeds the calibration threshold three times consecutively, the faulty sensor is automatically marked and a maintenance prompt is sent to the remote monitoring center; the improved ordinary Kriging interpolation algorithm is automatically used to complete the data in the monitoring area where the faulty sensor is located. The effective sensors within the corresponding radius are selected with the fault point as the center. The weight coefficients are calculated by modeling and solving the spherical variogram to calculate the completed value. Combined with time series trend correction and multi-model redundancy mechanism, the data completion error is ensured to be ≤10%, as shown in Tables 1 and 2.

[0027] The improved ordinary Kriging interpolation algorithm provided in this embodiment is an optimization and improvement based on the ordinary Kriging interpolation algorithm. The specific improvements are as follows: (1) Improved measurement point selection method: Based on the fault sensor, effective sensors are selected within a set radius to participate in the interpolation calculation, instead of using all measurement points in the entire domain, reducing redundant data, improving the computational efficiency in the edge computing environment, and meeting the real-time requirements of avalanche early warning. (2) Improved variogram model: A spherical variogram is used to model and solve the weight coefficients, replacing the traditional Gaussian model and exponential model, which is more in line with the spatial correlation distribution characteristics of avalanche monitoring data in high mountain and canyon areas, and improves the accuracy of interpolation weight calculation. (3) Improved time series trend correction: Based on spatial interpolation, a time series trend correction link is added, which integrates the historical change law of monitoring data into the interpolation calculation, overcomes the time series drift problem of meteorological, snow accumulation and vibration data in high altitude and large temperature difference environments, and improves data continuity. (4) Fault tolerance mechanism and accuracy improvement: A multi-model redundancy mechanism is introduced to perform multiple verifications and corrections on the interpolation results to ensure that the data completion error of the fault sensor area is ≤10% (as shown in Table 1 and Table 2), and the monitoring system can still operate stably and continuously under conditions such as sensor failure and data abnormality.

[0028] Table 1. Comparison of numerical completion errors between the improved ordinary Kriging interpolation algorithm and the ordinary Kriging interpolation algorithm in this embodiment.

[0029]

[0030] Table 2. Numerical completion error of the improved ordinary Kriging interpolation algorithm in this embodiment under different snow cover conditions.

[0031]

[0032] The distributed sensor unit includes: a meteorological sensor array, a seismic motion sensor array, a camera array, and a snow depth sensor array. Specifically: ① Meteorological sensors: measuring wind speed 0–60 m / s, wind direction 0–360°, air temperature -40–50°C, and snowfall 0–50 mm / h, with accuracies of ±0.1 m / s, ±1°, ±0.2°C, and ±0.1 mm / h respectively, deployed in the avalanche source area with a spacing ≤500 m. ② Seismic motion sensors: sensitivity ≥100 V / m / s, measurement frequency range 0.1–100 Hz, deployed in the flow area, forming a seismic network matrix with a spacing of 200–300 m. ③ High-definition camera array: pixels ≥4K, frame rate ≥30 fps, with night vision capability, night vision distance ≥100 m, deployed at a high altitude with a wide field of view, covering the avalanche initiation area, movement path, and accumulation area of ​​the target slope. ④ Snow depth sensor: Measurement range 0-10m, accuracy ±2cm, deployed at key sections in the avalanche source area, one unit every 100-150m. The key sections refer to slope sections within the avalanche source area with a slope range of 28° to 60°, possessing snow accumulation characteristics and prone to snow mass sliding initiation, including confluence channels, steep slope transition sections, ridge pass sections, and concave slope snow accumulation sections.

[0033] Figure 2 This graph illustrates the relationship between sensor deployment spacing (including meteorological sensors, seismic sensors, and snow depth sensors) and early warning response time. The horizontal axis represents sensor deployment spacing, ranging from 100 to 300 meters; the vertical axis represents early warning response time, including the total time for data acquisition, transmission delays, and model processing. Figure 2 It can be seen that a 200m deployment spacing corresponds to a response time of 8.3s, which meets the requirement of a high-level early warning response time of ≤10s; a 300m spacing corresponds to a response time of 12.5s, requiring the activation of the monitoring encryption mode; and a 150m spacing corresponds to a response time of 6.7s, which can be used as a reference for encryption deployment in high-risk areas. Therefore, the sensor deployment spacing in this embodiment balances monitoring cost and early warning efficiency.

[0034] In this embodiment, two sets of meteorological sensors and two sets of snow depth sensors are deployed in the avalanche source area above the tunnel entrance to monitor wind speed, wind direction, temperature, and snowfall; three sets of seismic sensors are deployed in the circulation area to monitor the vibration signal when the avalanche is initiated; and two sets of high-definition cameras with a resolution of ≥4K pixels and night vision capability are deployed at high, open locations on both sides of the tunnel entrance to collect real-time images of snow movement.

[0035] The tiered buffer protection module is deployed from top to bottom along the avalanche path, including a primary flow guiding structure, a secondary barrier structure, and a tertiary buffer structure. The primary flow guiding structure directs the avalanche flow towards a pre-designed channel and is located below the avalanche source area; the secondary barrier structure blocks the main snow mass and is located in the lower middle part of the flow area; the tertiary buffer structure weakens residual impact and is located between the construction camp and the secondary barrier structure. Specific parameters for each structure are as follows.

[0036] ① The primary diversion structure is an arc-shaped snow guide embankment. The arc-shaped snow guide embankment is made of C30 reinforced concrete and the surface is covered with a 100% solid content aromatic pure polyurea wear-resistant coating. It is formed by high pressure, high temperature and airless spraying. The coating thickness is ≥5mm, Shore D hardness is ≥45, Taber abrasion is ≤30mg / 500r, tensile strength is ≥16MPa and adhesion to concrete is ≥4.0MPa. The axis of the arc-shaped snow guide embankment is at an angle of 25° with the main avalanche flow direction, the radius of curvature is 50~80m, the height is 3m and the length is 50m.

[0037] ② The secondary retaining structure is a graded snow barrier constructed 50m below the snow guide dike, with a total height of 11m, a length of 132-150m, a top width of 5-6m, and a slope ratio of 1:2.3-1:2.5. Along the height of the dam, it is divided into a lower rigid dam section and an upper flexible protection section. The lower rigid dam section is constructed using tunnel spoil compaction, with a reinforced concrete facing on the outer side, bearing the main load-bearing, anti-sliding, and anti-overturning functions. In this embodiment, the lower rigid dam section is 8m high, using tunnel spoil compaction, and has a 50cm thick reinforced concrete facing on the outer side. The upper flexible protection section is a wire mesh protective net, used to buffer avalanche impact, reduce kinetic energy, and prevent snow from overturning. In this embodiment, the wire mesh size is 8cm. The upper and lower sections work together to bear the load, forming a graded snow barrier with graded load-bearing and graded energy dissipation.

[0038] ③ The three-level buffer structure consists of a foamed concrete buffer layer and an array of energy dissipation mounds; the foamed concrete buffer layer has a density of 300-500 kg / m³, a compressive strength ≥0.8 MPa, a thickness of 50-80 cm, and is cast with C25 concrete; the energy dissipation mounds are 2 m in diameter and 2 m in height, arranged in a quincunx pattern with a spacing of 5 m; the foamed concrete buffer layer and the array of energy dissipation mounds weaken the impact energy of residual snow and protect camp facilities.

[0039] like Figure 3 As shown, based on the height of the avalanche front : The total avalanche volume and the height of the snow barrier were calculated and fitted. Relationship curve: goodness of fit This indicates a logarithmic positive correlation between the total avalanche volume and the height of the snow barrier. The dashed line in the figure marks the dam height corresponding to an extreme avalanche of 15m. An 11m dam height can accommodate a 25×10 4 ~30×10 4 A typical extreme avalanche of m³. In the above formula, Indicates the width of the trench. Indicates the length of the avalanche path. Indicates the area of ​​the snow catchment area; Indicates the maximum snow depth. This represents the total volume of the avalanche.

[0040] In this embodiment, the secondary barrier structure of the graded buffer protection module also includes an adaptive adjustment component. This component dynamically adjusts based on real-time data from the multi-source monitoring and early warning module: when an avalanche velocity > 30 m / s and snow thickness > 3 m are detected, the inclination angle of the flexible protection section is adjusted to 15°–20° via a hydraulic drive device to enhance the impact buffering effect; when a wind direction shift > 10° is detected, the curvature of the arc-shaped snow guide embankment of the primary guide structure is finely adjusted via an electric push rod to maintain a guide angle ≤ 30°. The response time for these adjustments is ≤ 5 seconds, and the operating temperature range of the hydraulic drive device is -20 to 50°, making it suitable for high-altitude, low-temperature environments.

[0041] The intelligent emergency escape module includes an escape channel body, an emergency guidance device, and a life support unit. The escape channel body is used to connect the construction camp with the safe area of ​​the tunnel opening, the emergency guidance device is used to indicate the optimal escape route in real time, and the life support unit is used to provide emergency supplies in extreme environments.

[0042] The specific parameters of the intelligent emergency escape module are as follows: ① The escape passage body is a steel corrugated pipe structure, 170m long and 2m in diameter. LED emergency lighting fixtures are installed every 5m along the line. The inner wall of the passage is covered with a 5cm thick polyurethane insulation layer and a 1cm thick modified ultra-high molecular weight polyethylene impact lining. The modified ultra-high molecular weight polyethylene impact lining is based on ultra-high molecular weight polyethylene with a molecular weight of 3 million to 5 million, and is modified by adding halogen-free flame retardants and toughening agents. The impact strength of a simple supported beam without notches at -40℃ is ≥180kJ / m² and the thickness is 10±1mm.

[0043] ② The emergency guidance system includes 10 sets of audible and visual alarms and 5 sets of dynamic signs, which are installed inside the escape route. When the warning is activated, the audible and visual alarms automatically turn on, emitting a red warning light and a buzzer; the dynamic signs receive signals from the multi-source monitoring and warning module and adjust the escape direction in real time according to the avalanche impact location.

[0044] ③ Life support unit: All life support units are set up inside the escape passage, with one set every 10-15m. Each life support unit includes a portable oxygen supply device, a first aid kit and a Beidou-3 satellite communication terminal to ensure the breathing safety of personnel and external communication in extreme environments.

[0045] In this embodiment, the AI ​​early warning model built into the edge computing node is an avalanche risk classification and prediction model based on multi-source data fusion. This model is a fusion model of LightGBM gradient boosting tree and LSTM long short-term memory neural network, sequentially including a multi-source heterogeneous data preprocessing layer, a snow stability coefficient calculation unit, a wind speed change feature extraction unit, a vibration signal wavelet transform feature extraction unit, an environmental parameter compensation unit, a threshold screening and machine learning classification fusion judgment layer, and an early warning output layer. The training data sources for the model are: historical avalanche data for the past ten years, on-site measured snow thickness, density, wind speed, vibration, and topographic data from 2022 to 2024, elevation data of mountains surrounding tunnel entrances, and high-altitude environmental parameter data from 3478m to 4920m. The model is trained by combining supervised learning offline pre-training with online incremental updates, using historical avalanche events as labeled samples, optimizing parameters through 5-fold cross-validation, and then deploying the lightweight model on the edge computing node after compression and quantization.

[0046] The AI ​​early warning model built into the edge computing node is based on real-time monitoring of snow depth, wind speed, wind direction, temperature, seismic signal amplitude, snow motion image data, and terrain slope, altitude, and historical avalanche data. Through feature extraction, anomaly identification, and risk classification of snow stability coefficient, wind speed mutation threshold, and seismic signal amplitude, it completes avalanche risk assessment and outputs three-level early warning signals along with corresponding early warning response times and intelligent emergency escape module linkage control commands. The three-level early warning signals are: ① Low-level warning: Snow depth 30–50 cm, wind speed 10–15 m / s, seismic signal amplitude < 0.1 g, early warning response time ≤ 20 s. ② Medium-level warning: Snow depth 50–60 cm, wind speed 15–25 m / s, seismic signal amplitude 0.1–0.3 g, early warning response time ≤ 15 s. ③ High-level warning: Snow depth > 60 cm, wind speed > 25 m / s, seismic signal amplitude > 0.3 g, early warning response time ≤ 10 s.

[0047] When a low-level warning is issued, the multi-source monitoring and warning module encrypts the data acquisition frequency, and the sensor data acquisition frequency is increased from once / 5 minutes to once / 1 minute. When a medium-level warning is issued, outdoor construction work is shut down, emergency lighting in escape routes is turned on and emergency guidance devices are put into standby mode, and unnecessary power is cut off. When a high-level warning is issued, a full-site audible and visual alarm is triggered, and emergency broadcasts guide personnel to evacuate along escape routes. The evacuation command transmission delay is ≤3 seconds, and the remote monitoring center simultaneously reports to the emergency management department to activate the rescue plan.

[0048] In this embodiment, when the snow depth exceeds 50cm and the wind speed is ≥15m / s, the edge computing node outputs a medium-level warning, the remote monitoring center shuts down outdoor construction, and the emergency lighting in the escape route is turned on; when the ground motion sensor detects an avalanche vibration signal amplitude >0.3g, the system upgrades to a high-level warning, the audible and visual alarm is activated, and the emergency broadcast guides construction personnel to evacuate along the escape route. The entire evacuation time is ≤2min, the snow barrier successfully blocks the main snow body, the buffer structure weakens the residual impact, and there are no casualties or major property losses.

[0049] This embodiment integrates a multi-source monitoring and early warning module, a tiered buffer protection module, and an intelligent emergency escape module. It achieves precise avalanche early warning through multi-dimensional data fusion, strengthens impact protection using a tiered barrier structure, and ensures personnel safety through intelligent escape routes. This embodiment is suitable for the special terrain conditions of high mountains and canyons, significantly improving avalanche disaster prevention and control capabilities during the construction and operation phases of tunnel projects, reducing casualties and property losses, and is applicable to traffic engineering protection in high-altitude avalanche-prone areas.

[0050] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A composite avalanche prevention system for tunnel entrances in high mountain and canyon areas, characterized in that, This includes a multi-source monitoring and early warning module with coordinated operation, a tiered buffer protection module, and an intelligent emergency escape module; The multi-source monitoring and early warning module includes a distributed sensor unit, an edge computing node, and a remote monitoring center. The distributed sensor unit is deployed in three levels: avalanche source area, circulation area, and accumulation area, in the mountains surrounding the tunnel entrance and the construction area. The edge computing node has a built-in AI early warning model and is deployed at the tunnel site. The graded buffer protection module is deployed from top to bottom along the avalanche movement path, including a primary flow guiding structure, a secondary blocking structure and a tertiary buffer structure. The primary flow guiding structure is used to guide the avalanche to flow into the preset channel, the secondary blocking structure is used to block the main snow body, and the tertiary buffer structure is used to weaken the residual impact. The intelligent emergency escape module includes an escape channel body, an emergency guidance device, and a life support unit. The escape channel body is used to connect the construction camp with the safe area of ​​the tunnel opening. The emergency guidance device is used to indicate the optimal escape route in real time. The life support unit is used to provide emergency supplies. The distributed sensor unit includes: a meteorological sensor array deployed in the avalanche source area with a spacing of ≤500m; a seismic motion sensor array deployed in the flow area, forming a seismic network matrix with a spacing of 200-300m; a camera array with a field of view covering the avalanche initiation area, movement path, and accumulation area of ​​the target slope; and a snow depth sensor array deployed at key sections in the avalanche source area with a spacing of 100-150m. The key sections refer to slope sections within the avalanche source area with a slope range of 28° to 60°, which have snow accumulation characteristics and are prone to snow sliding initiation, including confluence channel sections, steep slope transition sections, ridge pass sections, and concave snow accumulation sections on the slope. The specific parameters of each structure in the graded buffer protection module are as follows: The primary diversion structure is an arc-shaped snow guide dike, with a 100% solid content aromatic pure polyurea wear-resistant coating applied to its surface. The coating thickness is ≥5mm, the angle between the axis of the arc-shaped snow guide dike and the main avalanche flow direction is ≤30°, the radius of curvature is 50-80m, the height is 3-5m, and the length is 50-100m; the secondary barrier structure is a graded snow barrier dam, with a total height of 11-15m, a dam length of 132-150m, a top width of 5-6m, and a dam slope ratio of 1:2.3-1:2.5; along the height of the dam, it is divided into a lower rigid dam body and an upper flexible protection section. The lower rigid dam body is constructed by layering and compacting tunnel spoil, with a reinforced concrete facing on the outside; the upper flexible protection section is a wire mesh protective net; the tertiary buffer structure is a foamed concrete buffer layer and an energy dissipation mound array; The secondary barrier structure also includes an adaptive adjustment component, which dynamically adjusts based on real-time data from the multi-source monitoring and early warning module: when the avalanche velocity is detected to be >30m / s and the snow thickness is detected to be >3m, the inclination angle of the flexible protection section is adjusted to 15° to 20° to enhance the impact buffering effect; when the wind direction shift is detected to be >10°, the curvature of the arc-shaped snow guide embankment of the primary guide structure is finely adjusted by an electric push rod to keep the guide angle ≤30°. The multi-source monitoring and early warning module also includes a sensor adaptive calibration unit. The sensor adaptive calibration unit has a built-in environmental compensation algorithm that automatically calibrates the data accuracy of meteorological sensors, seismic sensors and snow depth sensors based on environmental parameters at an altitude of 3478-4920m. The sensor adaptive calibration unit also includes a fault self-diagnosis function: when the sensor data deviation exceeds the calibration threshold three times consecutively, the faulty sensor is automatically marked and a maintenance prompt is sent to the remote monitoring center; the improved ordinary Kriging interpolation algorithm is automatically activated in the monitoring area where the faulty sensor is located to complete the data, and effective sensors within the corresponding radius are selected with the fault point as the center. The weight coefficients are calculated by modeling and solving the spherical variogram to calculate the completed value, and combined with time series trend correction and multi-model redundancy mechanism, the data completion error is ensured to be ≤10%; The AI ​​early warning model is an avalanche risk classification and prediction model based on multi-source data fusion. It is a fusion model of LightGBM and LSTM neural networks, which includes a multi-source heterogeneous data preprocessing layer, a snow accumulation stability coefficient calculation unit, a wind speed change feature extraction unit, a vibration signal wavelet transform feature extraction unit, an environmental parameter compensation unit, a threshold screening and machine learning classification fusion judgment layer, and an early warning output layer.

2. The composite avalanche prevention system for tunnel entrances in high mountain and canyon areas according to claim 1, characterized in that, The AI ​​early warning model, based on real-time monitoring of snow depth, wind speed, wind direction, temperature, seismic signal amplitude, snow motion image data, as well as terrain slope, altitude, and historical avalanche data, outputs three-level early warning signals along with corresponding early warning response times and intelligent emergency escape module linkage control commands. The three-level early warning signals include: Low-level warning: Snow depth 30-50cm, wind speed 10-15m / s, vibration signal amplitude <0.1g, warning response time ≤20s; Medium-level warning: snow depth 50-60cm, wind speed 15-25m / s, vibration signal amplitude 0.1-0.3g, warning response time ≤15s; High-level warning: snow depth > 60cm, wind speed > 25m / s, vibration signal amplitude > 0.3g, warning response time ≤ 10s.

3. The composite avalanche prevention system for tunnel entrances in high mountain and canyon areas according to claim 1, characterized in that, The specific parameters of the intelligent emergency escape module are as follows: The escape tunnel is constructed of corrugated steel pipe with an inner diameter of ≥2m and a wall thickness of ≥12mm. The inner wall of the tunnel is covered with a polyurethane insulation layer and a modified ultra-high molecular weight polyethylene impact-resistant lining. The tunnel length is ≤170m, and LED emergency lighting fixtures are installed along the route. The emergency guidance device includes an audible and visual alarm and a dynamic sign. When the warning is activated, the audible and visual alarm will automatically turn on, and the dynamic sign will display the optimal escape route in real time by receiving signals from the multi-source monitoring and warning module. Life support unit: One set is set up every 10-15m, including portable oxygen supply device, first aid kit and Beidou-3 satellite communication terminal.

Images